Single-Particle Characterizations of Ambient Aerosols during a Wintertime Pollution Episode in Nanning : Local Emissions vs . Regional Transport

Ambient aerosol during a heavily polluted episode in wintertime was characterized using real-time single particle aerosol mass spectrometry (SPAMS) in urban Nanning, a capital city in the Southwestern China. More than two million individual particles analyzed by SPAMS were classified into 8 major clusters based on the mass spectral patterns. A group of vanadium-rich particles were identified as the emissions from mining and smelting of vanadium mineral and were taken as markers of regional transported industrial emissions when air masses traversed northeast inland regions from Nanning. Number fractions of other industrially-emitted particles, including chromium-rich, elemental/organic carbon, organic carbon and fly ash, also increased during the regional transport events. During stagnant periods, local emissions sources, including vehicle exhaust (like Ca-EC particles) and local coal-fired power plants, contributed to the fine particles. During most of the sampling period, biomass burning particles produced by bagasse combustion were the most abundant, contributing ~25– 80% of the total classified particles. Our observations suggest that biomass burning particles derived from industrial heat and electricity cogeneration processes could have a significant impact on the urban air quality without proper emission controls.


INTRODUCTION
China's environmental degradation has accelerated with its economic growth in the past three decades (Liu, 2008).Industrial transfer from the eastern developed coastal areas to the less developed middle-western inland areas in China has been used to rapidly promote economic growth in the less developed inland regions since the 1990s, and the gap in GDP per capita between the coastal and inland regions has narrowed (Lemoine et al., 2015).Coupled with this accelerated industrialization, the inland regions have been suffering from serious air pollution because of large increases in industrial emissions of sulfur dioxide (SO 2 ), nitrogen oxides (NO x ), non-methane volatile organic compounds (NMVOCs) and black carbon (Qin and Xie, 2012;Kanada et al., 2013;Zhao et al., 2013;Qiu et al., 2014).In China, developed coastal regions outsource emissions to less developed inland regions through the import of energy and goods (Feng et al., 2013;Zhao et al., 2015).In the national west-to-east electricity transmission program, provinces such as Guangxi, Guizhou, Yunnan, Inner Mongolia and Shanxi installed significant coal fired capacity to provide electricity to eastern China since the 1980s (Ming et al., 2013).Inner Mongolia and Shanxi export 23% and 36% of the electricity they generate by coal combustion to the eastern provinces, respectively (Feng et al., 2013).The low value-added but high carbon-intensive goods imported from central and western provinces by the developed coastal areas contributed to about 80% of the carbon emission in China (Feng et al., 2013).
The frequent heavy pollution episodes in inland areas has attracted scientists' attention and several studies have been carried out, focusing on the aerosol chemistry and source apportionment in northwestern China (Cao et al., 2005;Cong et al., 2009;Shen et al., 2009;Yang et al., 2011).Wang et al. (2014) observed a PM 2.5 peak concentration of about 500 µg m -3 during an extremely polluted event in Xi'an and found that industrial sources contributed 58% of PM 2.5 mass.Aerosol composition (Meng et al., 2013) and contributions from long-range transport (Wang et al., 2015) were studied in the less-developed regions on the Qinghai-Tibet plateau.More recently, highly oxidized organic species were observed in the Qinghai-Tibet plateau which originated from inland China and/or locally intense photochemical and aqueousphase processing (Xu et al., 2015) .
Guangxi province in the southwestern part China has been viewed as one of the most scenic places in the country.The beautiful karst landform there attracts more than 30 million visitors from all over the world every year.Nanning, a medium scale city with a population of 7.3 million in 2014, is the capital of Guangxi province (www.gxtj.gov.cn).Nanning is also the economic and trade center for the bilateral cooperation between China and Southeast Asian countries.Since 2004, the China and Association of Southeast Asian Nations (ASEAN) Exposition is held in Nanning each year to promote economic development in the Southeast Asian region.Driven by the motivation of industrialization and urbanization, Nanning has rapidly increased in economic growth in recent years.With the increase in residents' average income, Nanning has become the city with the largest number of private automobiles in Guangxi, with 1.65 million vehicles in 2014 (www.gxtj.gov.cn).Recently, elevated PM 2.5 levels have become a major concern in urban Nanning.The average PM 2.5 concentration of 59 µg m -3 and 46 µg m -3 , respectively in 2013 and 2014, exceeded the value of national air quality standard in China (35 µg m -3 ).Furthermore, much higher mass concentrations of PM 2.5 were observed during wintertime (Fig. S1).However, few studies have been carried out on the atmospheric environment in Nanning to date.
Herein, we report the first real-time single-particle chemical composition and mixing state data from urban Nanning using a Single Particle Aerosol Mass Spectrometer (SPAMS).Based on the mass spectral patterns of individual particles, airborne particles were classified into different types to reveal their sources and evolution processes.Combined with an analysis of air parcel trajectories and meteorological conditions, typical particle types from regionally-transported industrial emissions as well as local emissions were identified.Temporal variations of total particulate mass loading and the number fraction of each particle type could provide valuable information on how local emissions and the regionally-transported industrial emissions affect the atmospheric environment of Nanning in winter.

Sampling Site and Period
The instruments were placed in a cabin on the top of the number two building in the Guangxi Academy of Environmental Sciences (22°48′N, 108°20′E) at urban Nanning (Fig. S2).With the mixture of residential, traffic and industrial emissions, this measurement site represented the urban area of Nanning.
The vacuum aerodynamic diameter (D va ) and chemical composition of individual particles were measured by SPAMS from January 14 to January 28, 2014.The size distribution of the ambient aerosol in the range of ~10-700 nm (mobility diameter) was measured by a scanning mobility particle sizer (SMPS, TSI 3936), with a time resolution of 5 min.Hourly average concentrations of PM 2.5 and PM 10 , gaseous pollutants (SO 2 , NO, NO 2 , O 3 and CO) and meteorological data (wind speed, wind direction, relative humidity (RH), temperature) were measured online at the Nanning environmental monitoring center about 2 km away from the sampling site.The temporal profile of temperature, relative humidity, SO 2 , CO, O 3 , NO and NO 2 concentration during the sampling period is shown in Fig. S3.

Instruments and Measurement
Single particle aerosol mass spectrometry (SPAMS, Hexin Analytical Instrument CO., Ltd., Guangzhou, China) was used to obtain individual particle chemical and size information in the range of ~0.2-2 µm.Particles were sampled from an inlet about 1 m above the roof of the building, about 15 meters above the ground, and then transferred to the SPAMS through a PM 2.5 cyclone and a 3 m long conductive rubber hose with 5mm inner diameter.A pump was used to pull air through the sampling system at 3 L min -1 , to keep the PM 2.5 separator running well and minimizing particle residence time in the sampling line.Aerosols were dried by diffusion drying tubes before they reached the instrument.Details about the operation of single-particle mass spectrometers are provided in previous publication (Li et al., 2011).Briefly, particles are efficiently drawn into the vacuum chamber through an aerodynamic focusing lens and accelerated to a terminal size-dependent velocity.Particle speed is measured by detection of light scattered as the particles pass through two 532 nm continuous lasers fixed at a set distance.Particle size is calculated from the measured speed by calibration with polystyrene latex spheres (PSLs) with known diameters.Sized particles are desorbed and ionized by a third laser (Q switched Nd:YAG laser, ~0.7 mJ pulse -1 , 266 nm) in the ion source region of a dual-polarity time-of flight mass spectrometer.Positive and negative ions are then accelerated and extracted in order to generate two mass spectra from each sampled particle.The exact m/z of each positive or negative ion was calibrated with known chemical compounds, for example NaI.Besides, the data collecting software of SPAMS was also used to calibrate mass spectral with more than six ions both in positive and negative mass spectral (i.e., K + (m/z 39), Na + (m/z 23), Pb + (m/z 206, 207 and 208), NO 3 -(m/z -62)，HSO 4 -(m/z -97) and CN -(m/z -26)).

Cluster Analysis
All single particle mass spectra at each m/z, diameter, date and identity number acquired by SPAMS were converted into a dataset using Yaada (version 2.11, www.yaada.org),a software toolkit in Matlab (version R2011a), as described previously (Song et al., 1999).Based on the similarities of the mass-to-charge ratio and the peak intensity, particles were clustered using the ART-2a method with a vigilance factor of 0.85, a learning rate of 0.05 and 20 iterations.Ultimately, the final clusters are determined by manually combining these results according to their mass spectral characteristics.

Air Quality and Meteorological Conditions
The measured particle mass concentration (Fig. 1(a)) illustrates that the concentrations of PM 2.5 and PM 10 were highly variable, consistent with the sources of air masses and variations in local wind direction and speed.Based on the mass concentration of PM 2.5 , air mass back trajectories, and the predominant local wind direction during the observation period, the sampling period is divided into four periods and the corresponding means of PM 2.5 , SO 2 , NO, NO 2 and O 3 concentration, wind speed, temperature and relative humidity (RH) are also listed in Table 1.Each of these periods is discussed in the following section.
The PM 2.5 concentration exhibited a diurnal pattern associated with the variation of local wind direction and wind speed from 00:00 on 14 January to 11:00 on 23 January, which was named Polluted Period 1.The diurnal trend of PM 2.5 was characterized by the anti-correlation with wind speed (Fig. S4).The minimum hourly mean PM 2.5 was 76 µg m -3 at 18:00, but increased rapidly and reached a maximum value of 133 µg m -3 with the average wind speed of 0.6 m s -1 at 5:00 and then the concentration of PM 2.5 was relatively stable until about 11:00 (Fig. S4).
From 19:00 each day to 11:00 the next day during Polluted Period 1, the corresponding local wind direction was predominantly northeasterly (Fig. 2(d)).Coincident with the local wind direction, the 36 h back trajectories of air masses at 100 m arrival height were mainly derived from the northeast of the sampling site (Fig. 2(a)).Back-trajectories were computed every day ending at 8:00 (Local time) in Nanning (GMT+8), using the Hybrid Single-Particle Lagrangian Integrated Trajectory Model (HYSPLIT) developed by NOAA/ARL (http://ready.arl.noaa.gov/HYSPLIT.php).We also analyzed daily 36 hour air mass back trajectories at 00:00, 6:00, 12:00, 18:00 (Local Time) at the levels of 100 m and 500 m during Clean Period1 and 2, Polluted Period 1 and 2, respectively (as shown in Figs.S5, S6 and S7).
The trajectories at 500 m were quite similar with those at 100 m in all the sampling periods.As shown in Fig. 2(a), the air masses arriving the sampling site travelled through Chongqing, Guizhou and Hunan provinces and then passed Liuzhou, Shanglin and some industrial sites on the northeast of Nanning, which are the important regions of power plants, mining, steel and smelting, etc. in Guangxi Province, bringing in significant amounts of particulate pollutants.In contrast with the overnight hours, the mass concentration of PM 2.5 in afternoon (from 11:00 to 19:00) during Polluted Period 1 were much lower, associated with the increase in the frequency of southeast and east-southeast wind with considerable increasing wind speed (Fig. 2(e)).
At midday on 21 January, a new particle formation event was identified based on the typical "banana" shape in the time-resolved size distribution.Meanwhile the mean diameter of particles increased consistently with the increase in the PM 2.5 mass concentration during the new particle formation event (Fig. 1(b)).During the Clean Period 1 and 2 (relatively, as listed in Table1), the original air masses arriving at the sampling site came from the South China Sea (Fig. 2(c)), bringing in relatively clean air, which contributed to the decrease in the mass concentration of PM 2.5 and PM 10 .
After the Clean Period1, the wind speed decreased.The average wind speed was 1 m s -1 during Polluted Period 2 (as shown in Table 1), resulting in stagnant conditions.Hence, the particle mass concentration of PM 2.5 increased dramatically in this period, with a maximum value of 205 µg m -3 .It is important to note that the high levels of air pollution with an hourly mean PM 2.5 concentration as high as about 170 µg m -3 lasted for 45 hours under the stagnant conditions, and ambient particles mass concentration decreased at 14:00 on 27 January.The sources and chemical characterization  of ambient aerosols during Polluted Period 2 and the differences between Polluted Period 1 and 2 are discussed in the following section.

Chemical Properties of Individual Particles Particle Classification
A total of 2 136 733 particles with both positive and negative ion spectra were chemically analyzed by SPAMS, accounting for ~30% of all sized particles during the whole sampling period.The correlation between total hit particle counts by SPAMS and the PM 2.5 mass concentrations was good during the whole sampling period (R = 0.72) (as shown in Fig. S8).The mixing state of ambient aerosols can provide information about the source, aging and chemical processes impacting the particles in the atmosphere.We classified 97.8% of all hit particles into 8 major groups according to their mass spectral patterns.Classified categories are named mainly according to their dominant mass spectral characteristics, as shown in Table 2.The characters of major classified particles are described in supplementary material.The average mass spectral patterns and size distributions of the major classes are shown in Figs.S9 and S10.

Particles during the Regional Transport Events
The mass concentration of PM 2.5 showed a periodic oscillation during Polluted Period 1, as illustrated in Fig. 1.The prevailing wind from northeast during the overnight period (~19:00-11:00 in Polluted Period 1) brought in particulate pollutants from the inland industrial regions, resulting in ~2-5 times higher concentrations of PM 2.5 compared to the afternoon time (~11:00-19:00) in Polluted Period 1.Here, we named these overnight periods (~19:00-11:00 in Polluted Period 1) with high PM concentrations as regional transport events.
(1) Vanadium-rich particles from mining and smelting of vanadium mineral During these regional transport events, a unique group of vanadium-rich particles among the metal type were identified.The vanadium particle cluster observed in this study is characteristic of a strong vanadium signature (V + , m/z 51; VO + , m/z 67) and minor peaks of K + (m/z 39), Na + (m/z 23), Fe + (m/z 56), Al + /C 2 H 3 + (m/z 27) in the positive mass spectra.The negative mass spectra include mainly sulfate (HSO 4 -, m/z 97), nitrate (NO 3 -, m/z 62), nitrite (NO 2 -, m/z 46), phosphate (PO 3 -, m/z 79), and lower intensity signal from SiO 3 -(m/z 76), as shown in Fig. 3(a).The strong vanadium signature has typically been identified as a marker of emissions from limited sources, due to the residual fuels used by ships and refinery containing high concentration of vanadium (Healy et al., 2009;Ault et al., 2010).
As shown in Fig. 3(b), the number concentration per hour of vanadium-rich particles peaked almost every morning from 6:00 to10:00 while the winds blew from the northeast during the regional transport events.In Fig. 2(a), back trajectories during the regional transport periods illustrate that the air masses arriving at the sampling site travelled through Shanglin County 120 kilometers northeast of Nanning, where the reserves of vanadium mineral are about 1.5 million tons, the third largest in China.Mining of vanadium mineral and smelting of vanadium pentoxide are important industries in this region.Fig. S11 showed the air mass back trajectories at 00:00, 6:00, 12:00, 18:00 (Local time) at the two levels (100m and 500m) on 22 Jan.2014 when the highest number of vanadium-rich particles were observed.The additional air mass back trajectories further confirmed our assignment of vanadium-rich particles.In contrast, few vanadium-rich particles were observed during Clean Period 1 and 2 even when the original air masses derived from the South China Sea and passed through ports before reaching urban Nanning (Figs. 2(c) and 3(b)), bringing in more sea salt aerosols (~6 times) than during Polluted Period 1.Hence, in this study, vanadium-rich particles observed at the sampling site were likely derived from mining of vanadium mineral and smelting of vanadium pentoxide, carried in the air masses which passed through Shanglin industrial region.Further, the appearance of nitric acid (H(NO 3 ) 2 -, m/z 125) (Zauscher et al., 2013) in Fig. 3(a) indicates that there is an excess of nitrate observed internally mixed with the vanadium particles, suggesting that those particles have undergone significant atmospheric aging during transport.
Additionally, we found a group of chromium-rich particles among the metal type, which correlated well with vanadiumrich particles in number (Fig. 3(b)).The chromium-rich particle is characterized by a dominant chromium signature (Cr + , m/z 52; CrO + , m/z 68) and lower intensities of Fe + (m/z 56), K + (m/z 39), Na + (m/z 23) and Ni + (m/z 58), and mixed with sulfate, nitrate, nitric acid and elemental carbon (Fig. S12).This particle type likely originated from steel manufacturing process emissions and/or heavy industry (refinery, coal mine, power stations) (Calvo et al., 2013).
(2) ECOC, OC and PAH particles from regional transport ECOC, OC and PAH particles accounted for 24.1%, 18.7% and 7.3%, respectively, of the total classified particles during regional transport events and were 1.9 times, 1.6 times and 1.2 times higher than in Polluted Period 2, respectively (Fig. S13).The particle number and number fraction of ECOC, OC and PAH groups obviously increased overnight (~19:00-11:00) in Polluted Period 1 (Figs. 4 and S14).Primary sources of ECOC, OC and PAH are the many combustion sources such as coal combustion and vehicle exhaust, etc.It is difficult to differentiate these emission sources based only on the overall mass spectral patterns.However, the variations of number fractions of some certain subclasses of these particles can potentially reveal their sources.
As a subclass of ECOC, the ECOC-Li particle type, which is characterized by the presence of lithium (Li + , m/z 7) in addition to peaks of elemental carbon (C + , C 2 + , C 3 + , C 4 + , C 5 + ) and organic carbon (C 2 H 3 + and C 2 H 3 O + ).The temporal trend of the number of particles in the ECOC-Li type has a positive correlation coefficient with fly ash (R = 0.5) during Polluted Period 1 (Table S1).Fly ash particles containing Al + (m/z 27) and SiO 3 -(m/z 76) are derived from coal combustion processes, as demonstrated by Ahn and Lee (2006).Furthermore, the number fraction of ECOC-Li and fly ash types in the total classified particles during regional transport events in Polluted Period 1 were 4.7 times and 1.7 times higher than Polluted Period 2, respectively, confirming the contributions from inland large-scale coal-fired power plants.It is interesting to note that fewer ECOC-Li and fly ash particles were observed in Polluted Period 2, but with better correlation (R = 0.9) in particle number (Table S2 ), suggesting contributions from local coal-fired power plants.

Local Emissions during the Stagnant Period
Starting from 18:00, 25 Jan, the wind speed dropped dramatically and resulted in an extreme stagnation event.The stagnation period lasted for 45 hours (Polluted Period 2) until 14:00 on the 27 th of Jan, with an average wind speed of 1 m s -1 , favoring the accumulation of locally emitted pollutants.
(1) Biomass burning particles from local bagasse combustion Biomass burning particles were the most prevalent particle type by number, making up 48%, on average, of the total classified particles during the whole sampling period.Notably during Polluted Period 2, biomass burning particles contributed 64% of the total classified particles on average, with an hourly maximum of 84% (Figs. 4 and S14).
The biomass burning particle cluster is characterized by a strong peak of K + (m/z 39 and 41) mixed with organic carbon -, m/z 46) and organic nitrogen fragments (CN - and CNO -) (Zauscher et al., 2013;Huo et al., 2015).The commonly-observed potassium chloride signal (K 2 Cl + , m/z 113) was only observed in a small fraction (3.4%) of the all observed biomass burning particles, due to the displacement of chloride by nitrate or sulfate, as reported by Zauscher et al. (2013).
Guangxi province is a major producer of sugar in China, where the major planting area and yield of sugarcane accounts for ~68% of total production in China since 2000.Eleven million tonnes of sugar cane (16% of the total production in Guangxi province in 2012) were produced in Nanning (Report on Guangxi sugar industry development, ecological environment protection and green transformation (in Chinese) by Guangxi Sugar Industry Association, November, 2014).Bagasse is the residue of from sugarcane processing after the juice is extracted from the sugarcane.Assuming that 280 kg of bagasse with moisture content of 50% is produced per tonne of sugarcane (Lopes Silva et al., 2014a), 3.16 million tonnes of bagasse were produced due to sugar processing.About 75% of the bagasse was combusted as a bio-fuel for heat and electricity cogeneration in urban Nanning (unpublished report).Bagasse burning industrial activities usually extend from November to April (the extraction season of sugarcane) and thus makes important contributions to the local particulate emissions during this period.Compared to coal and fuel oil mixtures, the combustion of mixtures of coal, fuel oil, and bagasse has a roughly equal particulate emission factor but lower emission factors of SO 2 and NO 2 (Turn et al., 2006).So far, no laboratory studies have been carried on the single particle mass spectrometry of bagasse burning.However, it is reasonable to believe that bagasse burning particles should share the similar common mass spectral patterns as other types of biomass burning including wood, straw and corn et al., based on previous studies on the burning of different type of biomass mass (Zauscher et al., 2013;Huo et al., 2015).On Jan. 26 and 27, the number fraction of biomass burning particles reached up to 80%.Fig. S15 showed the air mass back trajectories at 12:00 on 26 Jan. 2014 and 00:00 on 27 Jan.2014 (LT) at the two levels (100 m and 500 m) when the highest number of biomass burning particles were observed.Most of the air mass back trajectories shown in Fig. S15 were quite close to Nanning city, consistent with the stagnant meteorological condition.However, few wild fire activities were detected by MODIS along the air mass back trajectories (as shown in Fig. S16), suggesting that the contribution from wild fire activity emissions to the biomass burning particles observed at the sampling site was minor.The steady and extremely high number fraction of biomass burning particles observed in this study should be mainly attributed to the local bagasse burning in Nanning.
Sugarcane bagasse has the highest production among all the crops residues in the world (Bocci et al., 2009).It's also one of the most used biomass in energy generation (Lopes Silva et al., 2014).Despite being less efficient than gasification and pyrolysis, direct combustion is usually employed in the sugarcane industry because it requires much less investment and process control (Evans et al., 2010).Nowadays, using of biomass resources for energy generation at power/heat plants has attracted widespread interest because of limited fossil fuels.Our observations suggest that biomass burning particles derived from industrial heat and electricity cogeneration processes could have a significant impact on the urban air quality without proper emission controls.
(2) EC particles related to local emissions In our study, 110,507 EC particles were observed during the whole sampling period, accoutering for ~5% in the total classified particles.The predominant sources of EC are incomplete combustion processes of fossil fuels and bio-fuel, involving solid fuels for industrial and residential uses and vehicle exhaust, etc. (Ning and Sioutas, 2010;Bond et al., 2013).In order to probe local particle composition and sources, the mixing state of EC particles in urban Nanning was investigated.
The EC particles are characterized in SPAMS by intense elemental carbon ion peaks mainly including C, C 2 , C 3 , C 4 and C 5 in positive and/or negative mass spectra.In order to distinguish biomass burning sources discussed above, we classified the EC particles with none or lower intensity (relative intensity < 3%) of potassium (K + , m/z 39 and 41) in the positive mass spectra.In Fig. S10(b), the size distribution of EC particles is seen to have a bi-modal distribution, with one mode in the diameter range of ~200-450 nm and the larger mode in the diameter range of ~450-2000 nm.Similar observations were made in Paris (Healy et al., 2012), where EC particles in the smaller mode was attributed to local vehicle emissions, and the larger mode was attributed to aging processes.
To probe the size distribution and mixing state of EC particles, we use ternary plots to observe the relationship between mixing of secondary inorganic compounds and size of EC particles, using the ion peaks 97, 62 and 24 in negative mass spectra as markers for sulfate, nitrate and C 2 , respectively.Most of the EC particles are distributed in two regions (Fig. 5).Particles with diameters in the range of ~200-450 nm (blue and purple dots) concentrated in the region where the intensity of C 2 (m/z 24) is more than 25 percent with higher sulfate (< 75%) and lower nitrate intensity (< 20%).In contrast, particles with diameter in the range of ~450-2000 nm (green, yellow and red dots) are mainly found in the region with high intensity of nitrate (> 60%).
The average mass spectra reveals the differences between the larger and smaller EC particles (Fig. S17).Similar elemental carbon signals were observed in positive mass spectra of larger and smaller EC particles, whereas the larger EC particles (D va > 450 nm) had much more intense sulfate (HSO 4 -, m/z 97) and ammonium (NH 3 + and NH 4 + ) peaks than the smaller particles (D va ≤ 450 nm).The signals of nitrate (m/z 62) and nitrite (m/z 46) were also found in the larger EC particles (D va > 450 nm), but a few in the small EC particles (D va ≤ 450 nm).Apparently, the larger EC Fig. 5. Ternary plots of associations between mixing state and diameter of elemental carbon particles.The ion peaks 97, 62 and 24 in negative mass spectra as markers for sulfate, nitrate and C 2 , respectively.particles, which are internally mixed with secondary components, have undergone intensive atmospheric aging processes.It is very difficult to do source apportionment for these aged EC particles.However, the ageing processes could be very interesting.In Polluted Period 2, 98% of aged EC particles (D va > 450 nm) contained nitrate, suggesting that contribution from particulate nitrate formation to the particle growth.We also noticed that, among these aged EC particles, only 30% included NH 4 + signal, while nitric acid signal (H(NO 3 ) 2 -, m/z 125) which is a symbol of strong particle acidity, appeared in about 10% of the aged EC particles.The acidity of EC particles in Nanning needs to be further investigated in the future.
The smaller EC particles (D va ≤ 450 nm) were relatively fresh in our study.Among them, a group of Ca-EC particles (a subclass of EC particles), which is characterized by elemental carbon ion peaks and Ca + (m/z 40), CaO + (m/z 56) and CaOH + (m/z 57) in the positive mass spectrum, accounted for 38% in number.Ca-EC particles were identified as the local primary emissions of lubricating oil from vehicle exhaust (Dall'Osto and Harrison, 2012).Fig. S18 showed that the temporal variation of the particle number of Ca-EC type correlated with the mass concentration of NO (R = 0.6) during the whole sampling period.The Ca-EC particles exhibited pronounced diurnal variation, with a major peak during morning rush hours (7:00-9:00 LT) and high values during night hours (20:00-4:00 LT) as shown in Fig. S19.Nanning municipal government regulates that the heavily loaded diesel trucks can only go into downtown area from 23:00 to 6:00 next day.The nighttime high number counts of Ca-EC particles should be due to the overlap of evening traffic emission and the heavily loaded diesel truck emissions.The diurnal variation of Ca-EC consisted with the traffic flow of diesel engine vehicles and confirmed our assignment.During Polluted Period 2, better correlations of Ca-EC particles with mass concentrations of NO (R = 0.8), and CO (R = 0.8) were found, respectively.Apparently, local vehicle emission made a major contribution to the EC particles in Nanning.

CONCLUSIONS
With rapid industrialization and urbanization, the west part of China including city of Nanning is facing great challenges in air quality deterioration.In this work, results obtained by a real-time single particle analysis reveal the impact of local emissions and regional transport on the atmospheric environment of urban Nanning during heavily polluted wintertime.
Vanadium-rich particles were derived from mining of vanadium mineral and smelting of vanadium pentoxide, carried in the air masses which passed through the Shanglin industrial region.Particles attributed to the northeastern industrial region had a much higher number concentration of fine particles during regional transport events.During Polluted Period 2 when local emissions were dominant, fly ash and ECOC-Li particles were also observed, derived from primary local coal-fired power plants, and particles of Ca-EC from vehicle exhaust emissions also contributed to the fine particles.
During most of the sampling period, biomass burning particles produced by bagasse combustion made up the largest group and contributed 48% of the total classified particles with a high correlation with mass concentration of PM 2.5 during the whole sampling period, indicating a huge impact of sugar industry emissions on the air quality in Nanning.Considering the widespread interest of using biomass resources as a substitution of fossil fuels for energy generation, more strict emission controls should be applied to the direct combustion processes of biomass in industrial heat and electricity cogeneration.

Fig. 1 .
Fig. 1.(a) Time series of ambient wind speed, wind direction, PM 2.5 and PM 10 -PM 2.5 mass concentration.(b) Number distribution detected by SMPS and mean diameter over the sampling period.

Fig. 3 .
Fig. 3. (a) Average mass spectrum of the vanadium-rich type.(b) Temporal profile of vanadium-rich particle number (black solid line), chromium-rich (black dashed curve) and mass concentration of PM 2.5 (gray bar).The positive correlation coefficient between vanadium-rich and chromium-rich particles in number during the whole sampling periods was 0.67.

Table 1 .
The names, time, average of PM 2.5 , SO 2 , NO, NO 2 and O 3 concentration, Wind Speed (WS) , temperature (T) and relative humidity (RH) during the sampling period.

Table 2 .
Names and percent of the particle types identified from the SPAMS data.